Ad hoc classification of radiology reports.

OBJECTIVE: The task of ad hoc classification is to automatically place a large number of text documents into nonstandard categories that are determined by a user. The authors examine the use of statistical information retrieval techniques for ad hoc classification of dictated mammography reports. DESIGN: The authors' approach is the automated generation of a classification algorithm based on positive and negative evidence that is extracted from relevance-judged documents. Test documents are sorted into three conceptual bins: membership in a user-defined class, exclusion from the user-defined class, and uncertain. Documentation of absent findings through the use of negation and conjunction, a hallmark of interpretive test results, is managed by expansion and tokenization of these phrases. MEASUREMENTS: Classifier performance is evaluated using a single measure, the F measure, which provides a weighted combination of recall and precision of document sorting into true positive and true negative bins. RESULTS: Single terms are the most effective text feature in the classification profile, with some improvement provided by the addition of pairs of unordered terms to the profile. Excessive iterations of automated classifier enhancement degrade performance because of overtraining. Performance is best when the proportions of relevant and irrelevant documents in the training collection are close to equal. Special handling of negation phrases improves performance when the number of terms in the classification profile is limited. CONCLUSIONS: The ad hoc classifier system is a promising approach for the classification of large collections of medical documents. NegExpander can distinguish positive evidence from negative evidence when the negative evidence plays an important role in the classification.

Effective access to distributed heterogeneous medical text databases.

INQUERY is an advanced text information retrieval system developed by the Information Retrieval Laboratory of the University of Massachusetts in Amherst. It is based on Bayesian inference networks, which are probabilistic models for reasoning with multiple sources of uncertain evidence. The evidence, in this case, is the presence or absence of words and/or phrases in a document. Evidence is combined into belief that a document is relevant. The INQUERY retrieval engine has been developed with the support of ARPA, NSF, and industrial funding. It has a number of unique features and has achieved excellent results in the TIPSTER and TREC evaluations. Informatics research and application development using INQUERY has recently begun in the medical domain, including a new ARPA initiative concerned with clinical text. The features that we will focus on in this demonstration are: Automatic processing of natural language queries, including the extraction of phrases and specific medical concepts such as drug doses; Document selection through automatic relevance feedback and routing techniques, including the construction of complex queries using the INQUERY query language; The integration of conventional database techniques with text analysis and retrieval; Automatic thesaurus generation and query expansion using the PhraseFinder system; Distributed database access, including automatic database selection and merging of local searches; this will be demonstrated using a collection of medical databases; Retrieval based on passages, rather than whole documents.

Automated classification of encounter notes in a computer based medical record.

Harvard Community Health Plan is exploring emerging information technologies for means to use the text portion of its 25 year old computerized medical record system. The Center for Intelligent Information Retrieval is developing systems to answer the question: to what extent can automated information systems replace manual chart review of encounter notes? INQUERY, a probabilistic inference net information retrieval system, and FIGLEAF, an inductive decision tree text classifier are applied to the problem of classifying electronic encounter notes to identify acute exacerbations in pediatric asthmatics. Both systems achieve average precisions of greater than 80%, with a new enhancement to INQUERY's relevance feedback, the top performer. Refinement of the systems and plans for their integration are discussed.